NetCOLOR project
Last compiled on 22 October, 2021
This section shows some graphics about the processed data. The data consists in three .csv files:
metadata.csvhplc.csvabsorption.csv
sample_id 438116 was duplicated and removed from the data.
Notes
Will be using
depthinstead ofpressureto characterize the depth of the measurements.Will be only using observations measured in the first 2 meters of the water column. All the following graphs will be using data measured between 0-2 meters.
Mismatch between metadata, HPLC and absorption
Now, this table shows the metadata that have no HPLC values associated.
These are the absorption spectra with no HPLC association.
First visualizations
Most of the data is from the Scotian Shelf, Labrador Sea and Scotian Shelf and Slope.
The number of observation is also relatively stable through the years.
Bioregions
Observations were classified into bioregions based on the following criteria (where position is south if latitude < 48):
bathymetry > -300 & position == "North" ~ "Labrador & Greenland Shelves (LGS)"
bathymetry <= -300 & position == "North" ~ "Labrador Sea Basin (LSB)"
bathymetry >= -300 & position == "South" & yday <= 180 ~ "Scotian Shelf Spring (SSSp)"
bathymetry >= -300 & position == "South" & yday > 180 ~ "Scotian Shelf Fall (SSFa)"
bathymetry < -300 & position == "South"~ "Northwest Atlantic Basin ocean (NAB)"
HPLC
- I also noticed there are a lot of
0in the HPLC data. After discussions, they are true0and not missing values.
Some HPLC
sample_idwere not numeric (ex.: FL002), I removed them from the data.There are
hplchlaandhplcchlavariables in the data. I have merge both columns into one because I suspect it is an error and should be the same variable.but19was summed asbut19+butlike.hex19was summed ashex19+hexlike+hexlike2.I have calculated
aphy_specificusinghplcchla.Here is an attempt to visualize the seasonal cycle of few pigments. I have calculated the average pigment concentration and then the average total concentration to finally calculate the relative average contribution of each pigment.
Absorption
Here is how I named absorption data:
anapis the data contained in files withdetritusin their name.apis the data contained in files withparticulatein their name.aphyis the data contained in files withphytoplanktonin their name.
Absorption spectra with any values <= 0 between 350 and 400 nm have been removed.
Absorption spectra with
aphy(440)<aphy(410)have been removed (possible problem with pigment extraction).
ap
aphy
aphy specific
We can observe that there is a specific absorption gradient among bioregions.
anap
Calculating snap
The fit was done for data between 380 and 730 nm, excluding the 400–480 and 620–710 nm ranges to avoid any residual pigment absorption that might still have been present after sodium hypochlorite treatment (Babin 2003).
A R2 of 0.90 was used to filter out bad models.
Preliminary analysis
Seasonal cycle
- The next figure shows how many observations there are for each year of the day. We can see that there are temporal gaps in the data. For example, in
Labrador & Greenland Shelves (LGS), there are only data in May/June. It is not enough to get insights on the seasonal cycles. At the moment, I am not sure how to deal with that.
- A first attempt can be by using boxplots to see if there are differences among months and bioregions.
- Another way to visualize the same data.
- We can also look at the seasonal variability by bioregion. There are interesting patterns in the data. For instance, in
Northwest Atlantic Basin ocean (NAB)we can clearly see the seasonal pattern ofhex19.
Phytoplankton population model
Explore and try to understand the method of Emmanuel that aims to decompose absorption spectra into two populations.
Bricaud 1995
In this section, I put our data in the context of the paper by Bricaud et al. (1995).
- In table 2, they provide coefficients to fit \(a^*_\phi\) as a function of chlorophyll-a between 400 and 700 nm. The next figure shows some randomly selected spectral profiles of \(a^*_\phi\) and compare them with those modeled using the coefficients presented in Bricaud et al. (1995).
- The next figure shows \(a^*_\phi\) modeled at different wavelengths using the following equation:
\[ a^*_\phi(\lambda) = A(\lambda)\times\text{chl-a}^{-B(\lambda)} \]
Fucoxanthin bining
Relationships between fucoxanthin relative content and absorption
Based on the linear relationships in the previous graph, this shows the R2 as a function of the wavelength.
Apparent visible wavelength (AVW)
In this section we explore how the Apparent visible wavelength (AVW) could be used to describe the spectral shape of phytoplankton absorption. The AVW index is calculated following (Vandermeulen et al. 2020).
This graph shows all the phytoplankton absorption spectra colored by the value of the calculated AVW.
Figures for the manuscript
Figure 1
Figure 2
Attention: These numbers are based on the total number of reported stations. For example, in 2019, there are a total of 217 stations, but only 71 of them have absorption measurements.
Figure 3
There are some unusual bumps around 530 nm. We should have a closer look at the raw data.
Within each panel, we can see the averaged spectra and a subset of 100 randomly selected spectra by bioregion and type of absorption.
Figure 4
Figure 5
The observed nonlinearity between Chl_a and the ratio of phytoplankton absorption aph (443)/aph (670) indicating the packaging effect and changes in the intercellular composition of pigments.
Figure 6
Figure 7
Figure 8
Figure 9
Appendix
Appendix 1
Numbers at the right represent: mean (min - max).
Appendix 2
This aphy* ratio was found to correlate with phytoplankton cell size. That is cool to see that PAAW correlates with it! Using PAAW to infer aphy* means that we do not have to measure chla (HPLC or other ways).
